Ah, but that's not what is being attempted. Common Ancestors and Thrulines already try to do that, but with the possibility of errors introduced by incorrect information in the trees that are consulted.
Sideview doesn't actually work on a single match between two people. That single match will reveal commonality of one side of base pairs in only the specific segment lengths that match between two individuals. In itself, a single match is so small as to be insignificant, unless perhaps it involves a parent or sibling. But by comparing numerous matches with different individuals across different segments, it hopes to be able to phase your DNA to a lesser or greater extent by building a picture of the matching values known to be on one side in specific positions, and inferring the opposing values as a result.
Given enough pieces of the jigsaw, they hope to predictively phase SOME of your DNA, and yet other parts can be inferred to an extent in a similar way that partial completion of a 10,000 piece jigsaw puzzle may allow you to be confident that the picture is of red roses in a vase, even though you can't yet see them all.
And the software in itself cannot determine which side is maternal or paternal. It attempts to identify parent 1 and parent 2. It only gives you a maternal or paternal indication after you have allocated parent 1 and parent 2 accordingly. Get that wrong, and the whole thing falls over. Introduce the ability to compare your parental allocations with those of all the matches being used to attempt to phase your DNA, and the possibility of introducing errors by relying on decisions by random people rather than information from DNA samples that can be allocated and calculated, increases the likelihood of error exponentially.
It will probably never be perfect, and it probably works better for individuals with large numbers of matches on both sides of their family. But allowing it to make decisions based on how each match has allocated parent 1 and parent 2 (which may or may not always be correct) and how matches have been allocated to relationships in user trees (ditto) would just result in fuzzy information that wouldn't just create an error in that specific match between two individuals, but propagate across the entire attempt to phase your DNA to a lesser or greater degree.
It is just another tool which can recognise and calculate certain facts, from which it attempts to estimate and infer other parts that are missing. Given enough matches in certain positions within your DNA, it will likely become quite precise in those areas. With lower matches in other positions, there may be errors or an inability to allocate certain matches.
It's just another tool to aid your research, but as with all such tools, assertions need to be verified before you accept them as true or false. It may save you time in many cases by directing your research to one side of your tree when attempting to identify a match's relationship to you, and its aim is to be able to do that more often than not. There is still a possibility that it could send you up a creek without a paddle though
As Ancestry say, for 9 our of 10 people, the results are likely to be more than 95% accurate. To put it another way, if you are one of those 9 out of 10, then perhaps 1 in every 20 parental allocations in your list of DNA matches could be wrong. If you are the other 1 out of 10, it's worse than that.
